Posts

When starting with SharePoint development, a common question is “Which objects should I dispose?”. This is crucial because SharePoint leaks approximately 1 MB of memory per undisposed IDisposable object. For components like menus, this can quickly add up to 10 MB per page load. While the MSDN documentation provides comprehensive guidance, here’s the quick answer: In your WebPart, dispose of all SPWeb and SPSite objects except: SPContext.Current.Site SPContext.Current.Web SPContext.Current.Site.RootWeb For Features, dispose of all SPWeb and SPSite objects except those provided in your “properties” variable. ...

I edited this page on the 21 March 2010 because a lot of people seem interested and the code as since improved ! PHP has a pretty interesting feature, you can define a callback method to “catch” any error “thrown” in your code. And I’m sure most of you don’t use it. It’s really usefull when you want to make sure to detect error before any user reports it (which can takes time). This is all about avoiding to demolish with some lame errors your “user experience”. ...

Software version number is quite important. It helps you track what versions have your users when they report something. And when it’s linked to an SVN version number, it’s even better. Well, with MSBuild Community Task, you can easily automatically generate smart version numbers, you have to: Download MSBuildCommunityTasks Make sure your “svn.exe” binary is in C:\program files\subversion\bin Add this at the end of your .csproject file : 2011-07-02 update: As given in Markus comment, this code is a much better option: ...

MOVED HERE: javacint / FAQ I decided to open source the content I wrote in my blog pages to the javacint wiki. Please use it and participate from now on.

I’ve decided to start writing about my technical work and projects in English. While I’m comfortable reading English, I want to improve my writing skills. Feel free to provide feedback in the comments. Traduction : Je compte me remettre à écrire mais en anglais pour changer. blr_date: 2010-02-09 bw_readability_score_flesch: 76.5 bw_readability_score_coleman: 8.6 bw_readability_score_gunning: 8.3 bw_readability_score_smog: 6 bw_readability_score_ari: 4.6 bw_word_count: 93

To get the PL-2303 USB/Serial converter working on Vista, use the BAFO Technologies drivers available on their support site. Current version: Vista 32/64 v1.6

I recently rebuilt our TCP/UDP network server library to handle high-load scenarios. The initial thread-per-client approach worked but hit memory limits at around 3,000 simultaneous connections. The solution? Kernel events. Using IOCP (Windows), epoll (Linux), and kqueue (FreeBSD), we let the kernel notify our program when new data arrives instead of polling. With a 128-byte pre-allocated buffer and some BeginReceive calls, the changes were minimal. The results exceeded expectations: from a 10,000 connection target, we achieved 60,000 simultaneous connections with near-zero CPU usage and instant response times. This proves Mono/.NET’s capability for high-performance network servers.

I recently discovered an interesting mesh networking project called [B.A.T.M.A.N.][1] (Better Approach To Mobile Ad-hoc Networking). It builds upon and improves the [OLSR][2] protocol specifications for routing in mobile ad-hoc networks. You can learn more about this project on the [B.A.T.M.A.N. website][3].

When migrating a MapPoint WebService application from Windows to Mono/Linux, I encountered “401 Unauthorized” errors. Network analysis revealed slight differences in HTTP headers. Here’s the solution using specific credential handling: var cred = new System.Net.NetworkCredential("---user---", "---password---"); _credCache = new CredentialCache(); _credCache.Add( new Uri( "http://findv3.staging.mappoint.net" ), "Digest", cred ); _finder = new FindServiceSoap(); _finder.Credentials = _credCache; _finder.PreAuthenticate = true;

Today, I optimized a MySQL query to find spatial points by proximity. Initially, my query took an average of 250 ms. After some research, I came across a document from MySQL. By following their recommendation to use a preliminary approximation via a bounding box, I was able to reduce the query time to a range of 5 to 35 ms. However, as I continued reading, I discovered that MySQL also offers a spatial extension based on R-Tree. I thought I had found the perfect solution. But no, R-Trees are only available with the MyISAM engine, and creating spatial indexes is not allowed in InnoDB. The same goes for FULLTEXT indexes, which are also not available in InnoDB. ...